Rendered from docs/probe-results/campaign-of-2026-10-03-campaign-tier-mcp-arm-navigability.md in the Headwater corpus. Every document on this half of the site is typed by the taxonomy the descriptor names: corpus.json.

The result of docs/probe-runs/campaign-of-2026-10-03-campaign-tier-mcp-arm-navigability.md

A probe result is a function of committed inputs and of nothing else: the transcript at docs/probe-runs/campaign-of-2026-10-03-campaign-tier-mcp-arm-navigability.md and the state it stands at, the expectations the probes it names declare, the version of the grader that evaluated them, and for a compared arm the result of the other arm. Fetch those and this file comes back.

The run this transcript recorded

A campaign run in the mcp arm, on claude-sonnet-5 at 2026-10-03. served version claude-sonnet-5, tree sha256:81d5f62acb289022175d192df0a61faf5f71eb4b79e8dd1f4efae4532ebdd916, selection sha256:77ce7e13166fd99f67da4bb4ddd90eaa6076dedfb335bc54563f94dce33b0238, read set sha256:18ae82cb33e60fa707b36ccc3090f962d3b82ac1a179359fdaecfc1a30e6be7b, seed 0, harness 0.5.0. realized cost $29.10, which the adaptive layer reads as the cost of its own instrument.

150 events over 5 probes, in 150 sessions and 632 tool calls.

The engine confirmed the taxonomy, that every member of the run identity is present, the membership of every probe named, that no key outside the closed set appears, and that a realized cost was recorded. Present is not confirmed. This page names the lock, the selection and the read set the transcript recorded, and it compares none of them with the tree in front of a reader. headwater generate and headwater probe stale name what moved since the recording, and neither one fails for it. It graded nothing: a verdict is a function of this transcript, the expectations these probes declare and a grader version, and headwater probe grade is the verb that holds all three.

The selection this transcript names is sha256:77ce7e13166fd99f67da4bb4ddd90eaa6076dedfb335bc54563f94dce33b0238, and the probes graded below are the probes it was planned over, as this corpus declares them now. A probe added to this corpus after the recording is not a session this run owed. The tree, the seed and the harness above are provenance and nothing compares them.

The read_set digest above covers every probe of the selection and every document one of them examines, by path and content. This file names it as provenance and compares it with nothing, and the same holds for the lock and the selection: an edit to a document, a new probe or a moved lock leaves these bytes alone. headwater generate names each result whose recorded lock, read set or selection moved, and headwater probe stale takes the digest and reports which recorded results a change voided. Neither one fails for it.

Graded by grader 0.5.0. A campaign run in the mcp arm, on claude-sonnet-5 at served version claude-sonnet-5.

The verdicts

  • HW-PROBE-a-session-answers-from-the-register-without-opening-the-question-it-replaced (navigability, expects not_opened) session L20-campaign-mcp-p2-r1: satisfied — 4 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r10: satisfied — 6 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r11: satisfied — 5 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r12: satisfied — 7 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r13: satisfied — 9 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r14: satisfied — 5 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r15: satisfied — 12 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r16: satisfied — 6 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r17: satisfied — 5 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r18: satisfied — 5 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r19: satisfied — 6 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r2: satisfied — 6 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r20: satisfied — 8 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r21: satisfied — 4 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r22: satisfied — 6 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r23: satisfied — 3 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r24: satisfied — 4 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r25: satisfied — 5 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r26: satisfied — 6 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r27: satisfied — 6 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r28: satisfied — 8 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r29: satisfied — 5 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r3: satisfied — 9 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r30: satisfied — 4 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r4: satisfied — 4 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r5: satisfied — 5 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r6: satisfied — 6 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r7: satisfied — 4 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r8: satisfied — 4 recorded calls, and none named any of the 1 document session L20-campaign-mcp-p2-r9: satisfied — 6 recorded calls, and none named any of the 1 document
  • HW-PROBE-a-session-says-how-a-governs-line-reaches-a-document (navigability, expects answered) session L20-campaign-mcp-p3-r1: not satisfied — event 61 answered hand, and the probe expects command session L20-campaign-mcp-p3-r10: not satisfied — event 62 answered hand, and the probe expects command session L20-campaign-mcp-p3-r11: not satisfied — event 63 answered hand, and the probe expects command session L20-campaign-mcp-p3-r12: not satisfied — event 64 answered hand, and the probe expects command session L20-campaign-mcp-p3-r13: not satisfied — event 65 answered hand, and the probe expects command session L20-campaign-mcp-p3-r14: not satisfied — event 66 answered hand, and the probe expects command session L20-campaign-mcp-p3-r15: not satisfied — event 67 answered hand, and the probe expects command session L20-campaign-mcp-p3-r16: not satisfied — event 68 answered hand, and the probe expects command session L20-campaign-mcp-p3-r17: not satisfied — event 69 answered hand, and the probe expects command session L20-campaign-mcp-p3-r18: not satisfied — event 70 answered hand, and the probe expects command session L20-campaign-mcp-p3-r19: not satisfied — event 71 answered hand, and the probe expects command session L20-campaign-mcp-p3-r2: not satisfied — event 72 answered hand, and the probe expects command session L20-campaign-mcp-p3-r20: not satisfied — event 73 answered hand, and the probe expects command session L20-campaign-mcp-p3-r21: not satisfied — event 74 answered hand, and the probe expects command session L20-campaign-mcp-p3-r22: not satisfied — event 75 answered hand, and the probe expects command session L20-campaign-mcp-p3-r23: not satisfied — event 76 answered hand, and the probe expects command session L20-campaign-mcp-p3-r24: not satisfied — event 77 answered hand, and the probe expects command session L20-campaign-mcp-p3-r25: not satisfied — event 78 answered hand, and the probe expects command session L20-campaign-mcp-p3-r26: not satisfied — event 79 answered hand, and the probe expects command session L20-campaign-mcp-p3-r27: not satisfied — event 80 answered hand, and the probe expects command session L20-campaign-mcp-p3-r28: not satisfied — event 81 answered hand, and the probe expects command session L20-campaign-mcp-p3-r29: not satisfied — event 82 answered hand, and the probe expects command session L20-campaign-mcp-p3-r3: not satisfied — event 83 answered hand, and the probe expects command session L20-campaign-mcp-p3-r30: not satisfied — event 84 answered hand, and the probe expects command session L20-campaign-mcp-p3-r4: not satisfied — event 85 answered hand, and the probe expects command session L20-campaign-mcp-p3-r5: not satisfied — event 86 answered hand, and the probe expects command session L20-campaign-mcp-p3-r6: not satisfied — event 87 answered hand, and the probe expects command session L20-campaign-mcp-p3-r7: not satisfied — event 88 answered hand, and the probe expects command session L20-campaign-mcp-p3-r8: not satisfied — event 89 answered hand, and the probe expects command session L20-campaign-mcp-p3-r9: satisfied — event 90: answered command
  • HW-PROBE-a-session-says-where-a-figure-on-the-site-is-written (navigability, expects answered) session L20-campaign-mcp-p4-r1: satisfied — event 91: answered published session L20-campaign-mcp-p4-r10: satisfied — event 92: answered published session L20-campaign-mcp-p4-r11: satisfied — event 93: answered published session L20-campaign-mcp-p4-r12: satisfied — event 94: answered published session L20-campaign-mcp-p4-r13: satisfied — event 95: answered published session L20-campaign-mcp-p4-r14: satisfied — event 96: answered published session L20-campaign-mcp-p4-r15: satisfied — event 97: answered published session L20-campaign-mcp-p4-r16: satisfied — event 98: answered published session L20-campaign-mcp-p4-r17: satisfied — event 99: answered published session L20-campaign-mcp-p4-r18: satisfied — event 100: answered published session L20-campaign-mcp-p4-r19: satisfied — event 101: answered published session L20-campaign-mcp-p4-r2: satisfied — event 102: answered published session L20-campaign-mcp-p4-r20: satisfied — event 103: answered published session L20-campaign-mcp-p4-r21: satisfied — event 104: answered published session L20-campaign-mcp-p4-r22: satisfied — event 105: answered published session L20-campaign-mcp-p4-r23: satisfied — event 106: answered published session L20-campaign-mcp-p4-r24: satisfied — event 107: answered published session L20-campaign-mcp-p4-r25: satisfied — event 108: answered published session L20-campaign-mcp-p4-r26: satisfied — event 109: answered published session L20-campaign-mcp-p4-r27: satisfied — event 110: answered published session L20-campaign-mcp-p4-r28: satisfied — event 111: answered published session L20-campaign-mcp-p4-r29: satisfied — event 112: answered published session L20-campaign-mcp-p4-r3: satisfied — event 113: answered published session L20-campaign-mcp-p4-r30: satisfied — event 114: answered published session L20-campaign-mcp-p4-r4: satisfied — event 115: answered published session L20-campaign-mcp-p4-r5: satisfied — event 116: answered published session L20-campaign-mcp-p4-r6: satisfied — event 117: answered published session L20-campaign-mcp-p4-r7: satisfied — event 118: answered published session L20-campaign-mcp-p4-r8: satisfied — event 119: answered published session L20-campaign-mcp-p4-r9: satisfied — event 120: answered published
  • HW-PROBE-an-agent-reaches-the-adjudication-from-the-document-that-lost-it (navigability, expects opened) session L20-campaign-mcp-p1-r1: satisfied — event 1, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r1/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r10: satisfied — event 2, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r10/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r11: satisfied — event 3, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r11/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r12: satisfied — event 4, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r12/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r13: satisfied — event 5, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r13/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r14: satisfied — event 6, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r14/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r15: satisfied — event 7, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r15/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r16: satisfied — event 8, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r16/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r17: satisfied — event 9, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r17/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r18: satisfied — event 10, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r18/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r19: satisfied — event 11, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r19/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r2: satisfied — event 12, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r2/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r20: satisfied — event 13, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r20/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r21: satisfied — event 14, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r21/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r22: satisfied — event 15, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r22/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r23: satisfied — event 16, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r23/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r24: satisfied — event 17, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r24/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r25: satisfied — event 18, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r25/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r26: satisfied — event 19, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r26/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r27: satisfied — event 20, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r27/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r28: satisfied — event 21, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r28/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r29: satisfied — event 22, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r29/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r3: satisfied — event 23, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r3/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r30: satisfied — event 24, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r30/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r4: satisfied — event 25, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r4/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r5: satisfied — event 26, call 4: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r5/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r6: satisfied — event 27, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r6/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r7: satisfied — event 28, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r7/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r8: satisfied — event 29, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r8/docs/decisions/0008-probe-cost-and-cadence.md session L20-campaign-mcp-p1-r9: satisfied — event 30, call 3: Read /var/tmp/hw-1659/a/ws/L20-campaign-mcp-p1-r9/docs/decisions/0008-probe-cost-and-cadence.md
  • HW-PROBE-what-a-session-writes-points-back-at-the-ruling-it-rests-on (navigability, expects cited) session L20-campaign-mcp-p5-r1: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r10: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r11: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r12: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r13: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r14: satisfied — event 126: docs/explanations/a-corpus-wide-count-is-computed-when-it-is-read.md cites HW-DR-0049 session L20-campaign-mcp-p5-r15: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r16: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r17: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r18: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r19: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r2: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r20: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r21: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r22: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r23: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r24: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r25: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r26: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r27: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r28: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r29: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r3: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r30: satisfied — event 144: engine/crates/census/src/census.rs cites HW-DR-0049 session L20-campaign-mcp-p5-r4: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r5: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r6: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r7: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r8: not satisfied — 0 produced artifacts, and none cites any of the 1 identifier session L20-campaign-mcp-p5-r9: satisfied — event 150: engine/crates/census/src/census.rs cites HW-DR-0049

The rate, and the denominator it is over

94 of 150 graded sessions satisfied their expectation: 62.7%, in a 95% interval of 54.7% to 70.0%. The denominator is the graded sessions and never the selected probes. 0 sessions reached no verdict, and a session with no verdict is outside both halves of that fraction.

An interval that overlaps the previous run's is variance and one that does not is drift. This is one arm, so it estimates no effect: an efficacy claim is a comparison of two results, and the arm each one recorded is on it.

The comparisons this arm takes part in

What the MCP server adds. The treated arm is the campaign mcp arm, which added the paths the mcp component's delta names, and the control is the campaign present arm. The documents are in both arms (spec 5).

  • treated, docs/probe-runs/campaign-of-2026-10-03-campaign-tier-mcp-arm-navigability.md: 94 of 150 graded sessions satisfied their expectation, 62.7%, in a 95% interval of 54.7% to 70.0%. 0 sessions refused by the session itself.
  • control, docs/probe-runs/campaign-of-2026-10-03-campaign-tier-present-arm-navigability.md: 95 of 150 graded sessions satisfied their expectation, 63.3%, in a 95% interval of 55.4% to 70.6%. 0 sessions refused by the session itself.

The difference is -0.7 points, in a 95% Newcombe interval of -11.5 points to +10.2 points. The interval contains zero, so this run does not separate the two arms at the 5% level.